Définition
Acoustic modeling underlies systems that turn speech into text. The model analyzes frequencies, pauses, intonation, noise and pronunciation features, and then estimates which speech units were spoken. Modern systems often combine the acoustic part with a language model to select a more natural version of the phrase.
Exemple
When a person says “put the meeting at five,” the speech recognition system must distinguish the command from similar-sounding words and take into account the context.
Pourquoi c'est important
The term is important for choosing transcription services, voice assistants, subtitling, call centers and call analysis.
Fonctionnement
The sound is broken into short fragments, features are extracted and fed into the model. It estimates the probabilities of sound sequences, and the system assembles them into words, taking into account the dictionary and context.
Où c'est utilisé
- speech recognition
- automatic subtitles
- voice assistants and call analytics
Limites
The quality decreases with strong noise, poor microphone, rare accents, fast speech and specialized vocabulary. Business often requires adaptation to the domain.
FAQ
Why is “Acoustic Modeling” useful to know?
The term is important for choosing transcription services, voice assistants, subtitling, call centers and call analysis.